def Net(dataReader, num_input, num_hidden, num_output, params):
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Relu())
net.add_layer(relu1, "relu1")
fc2 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc2, "fc2")
relu2 = ActivatorLayer(Relu())
net.add_layer(relu2, "relu2")
fc3 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc3, "fc3")
relu3 = ActivatorLayer(Relu())
net.add_layer(relu3, "relu3")
fc4 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc4, "fc4")
relu4 = ActivatorLayer(Relu())
net.add_layer(relu4, "relu4")
fc5 = FcLayer(num_hidden, num_output, params)
net.add_layer(fc5, "fc5")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1, need_test=False)
net.ShowLossHistory()
def Model(dataReader, num_input, num_hidden, num_output, params):
net = NeuralNet41(params, "overfitting")
fc1 = FcLayer(num_input, num_hidden, params)
net.add_layer(fc1, "fc1")
s1 = ActivatorLayer(Sigmoid())
net.add_layer(s1, "s1")
fc2 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc2, "fc2")
t2 = ActivatorLayer(Tanh())
net.add_layer(t2, "t2")
fc3 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc3, "fc3")
t3 = ActivatorLayer(Tanh())
net.add_layer(t3, "t3")
fc4 = FcLayer(num_hidden, num_output, params)
net.add_layer(fc4, "fc4")
net.train(dataReader, checkpoint=100, need_test=True)
net.ShowLossHistory(XCoordinate.Epoch, ymin=0.8)
return net
def DropoutNet(dataReader, num_input, num_hidden, num_output, params):
net = NeuralNet(params)
fc1 = FcLayer(784, 128, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Relu())
net.add_layer(relu1, "relu1")
drop1 = DropoutLayer(128, 0.3)
net.add_layer(drop1, "dp1")
fc2 = FcLayer(128, 32, params)
net.add_layer(fc2, "fc2")
relu2 = ActivatorLayer(Relu())
net.add_layer(relu2, "relu2")
drop2 = DropoutLayer(32, 0.5)
net.add_layer(drop2, "dp2")
fc2 = FcLayer(32, 16, params)
net.add_layer(fc2, "fc2")
relu2 = ActivatorLayer(Relu())
net.add_layer(relu2, "relu2")
drop2 = DropoutLayer(16, 0.5)
net.add_layer(drop2, "dp2")
fc5 = FcLayer(16, 10, params)
net.add_layer(fc5, "fc5")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1)
net.ShowLossHistory()
def DropoutNet(num_input, num_hidden1, num_hidden2, num_output, params):
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid = ActivatorLayer(Relu())
net.add_layer(sigmoid, "sigmoid")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
tanh = ActivatorLayer(Relu())
net.add_layer(tanh, "tanh")
dp1 = DropoutLayer(num_hidden2, 0.5)
net.add_layer(dp1, "dp1")
fc3 = FcLayer(num_hidden2, num_output, params)
net.add_layer(fc3, "fc3")
dp2 = DropoutLayer(num_output, 0.5)
net.add_layer(dp2, "dp2")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=10, need_test=True)
net.ShowLossHistory(0, None, 0, 1)
def model():
dr = LoadData()
num_input = dr.num_feature
num_hidden1 = 64
num_hidden2 = 64
num_hidden3 = 32
num_hidden4 = 16
num_output = 1
max_epoch = 100
batch_size = 16
learning_rate = 0.1
eps = 1e-3
params = HyperParameters40(learning_rate,
max_epoch,
batch_size,
eps,
net_type=NetType.BinaryClassifier,
init_method=InitialMethod.Xavier)
net = NeuralNet40(params, "Income")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
a1 = ActivatorLayer(Relu())
net.add_layer(a1, "relu1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
a2 = ActivatorLayer(Relu())
net.add_layer(a2, "relu2")
fc3 = FcLayer(num_hidden2, num_hidden3, params)
net.add_layer(fc3, "fc3")
a3 = ActivatorLayer(Relu())
net.add_layer(a3, "relu3")
fc4 = FcLayer(num_hidden3, num_hidden4, params)
net.add_layer(fc4, "fc4")
a4 = ActivatorLayer(Relu())
net.add_layer(a4, "relu4")
fc5 = FcLayer(num_hidden4, num_output, params)
net.add_layer(fc5, "fc5")
logistic = ClassificationLayer(Logistic())
net.add_layer(logistic, "logistic")
#net.load_parameters()
net.train(dr, checkpoint=10, need_test=True)
net.ShowLossHistory("epoch")
def model2():
dr = LoadData()
num_input = dr.num_feature
num_hidden1 = 64
num_hidden2 = 64
num_hidden3 = 32
num_hidden4 = 16
num_output = 1
max_epoch = 1000
batch_size = 16
learning_rate = 0.01
eps = 0.001
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy2, InitialMethod.Xavier,
OptimizerName.Adam)
net = NeuralNet(params, "Income")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
a1 = ActivatorLayer(Relu())
net.add_layer(a1, "relu1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
a2 = ActivatorLayer(Relu())
net.add_layer(a2, "relu2")
fc3 = FcLayer(num_hidden2, num_hidden3, params)
net.add_layer(fc3, "fc3")
a3 = ActivatorLayer(Relu())
net.add_layer(a3, "relu3")
fc4 = FcLayer(num_hidden3, num_hidden4, params)
net.add_layer(fc4, "fc4")
a4 = ActivatorLayer(Relu())
net.add_layer(a4, "relu4")
fc5 = FcLayer(num_hidden4, num_output, params)
net.add_layer(fc5, "fc5")
sigmoid5 = ClassificationLayer(Sigmoid())
net.add_layer(sigmoid5, "sigmoid5")
#net.load_parameters()
net.train(dr, checkpoint=10, need_test=True)
net.ShowLossHistory()
def model():
dataReader = LoadData()
num_input = 1
num_hidden1 = 4
num_output = 1
max_epoch = 10000
batch_size = 10
learning_rate = 0.5
eps = 0.001
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.MSE, InitialMethod.Xavier,
OptimizerName.SGD)
net = NeuralNet(params, "Level1_CurveFittingNet")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid1 = ActivatorLayer(Sigmoid())
net.add_layer(sigmoid1, "sigmoid1")
fc2 = FcLayer(num_hidden1, num_output, params)
net.add_layer(fc2, "fc2")
#net.load_parameters()
#ShowResult(net, dataReader, params.toString())
#ShowResult2(net, dataReader)
net.train(dataReader, checkpoint=100, need_test=True)
net.ShowLossHistory()
#ShowResult(net, dataReader, params.toString())
ShowResult(net, dataReader)
def model():
dataReader = LoadData()
num_input = 1
num_hidden1 = 4
num_output = 1
max_epoch = 10000
batch_size = 10
learning_rate = 0.5
eps = 1e-5
params = HyperParameters40(learning_rate,
max_epoch,
batch_size,
net_type=NetType.Fitting,
init_method=InitialMethod.Xavier)
net = NeuralNet40(params, "Level1_CurveFittingNet")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid1 = ActivatorLayer(Sigmoid())
net.add_layer(sigmoid1, "sigmoid1")
fc2 = FcLayer(num_hidden1, num_output, params)
net.add_layer(fc2, "fc2")
net.train(dataReader, checkpoint=100, need_test=True)
net.ShowLossHistory("epoch")
ShowResult(net, dataReader)
def model():
dataReader = LoadData()
num_input = dataReader.num_feature
num_hidden1 = 8
num_output = 3
max_epoch = 5000
batch_size = 10
learning_rate = 0.1
eps = 0.06
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy3, InitialMethod.Xavier,
OptimizerName.SGD)
net = NeuralNet(params, "chinabank")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
r1 = ActivatorLayer(Relu())
net.add_layer(r1, "Relu1")
fc2 = FcLayer(num_hidden1, num_output, params)
net.add_layer(fc2, "fc2")
softmax1 = ClassificationLayer(Softmax())
net.add_layer(softmax1, "softmax1")
net.train(dataReader, checkpoint=10, need_test=True)
net.ShowLossHistory()
ShowResult(net, params.toString())
ShowData(dataReader)
def DropoutNet(dataReader, num_input, num_hidden, num_output, params):
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Relu())
net.add_layer(relu1, "relu1")
drop1 = DropoutLayer(num_hidden, 0.1)
net.add_layer(drop1, "dp1")
fc2 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc2, "fc2")
relu2 = ActivatorLayer(Relu())
net.add_layer(relu2, "relu2")
drop2 = DropoutLayer(num_hidden, 0.3)
net.add_layer(drop2, "dp2")
fc3 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc3, "fc3")
relu3 = ActivatorLayer(Relu())
net.add_layer(relu3, "relu3")
drop3 = DropoutLayer(num_hidden, 0.3)
net.add_layer(drop1, "dp3")
fc4 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc4, "fc4")
relu4 = ActivatorLayer(Relu())
net.add_layer(relu4, "relu4")
drop4 = DropoutLayer(num_hidden, 0.3)
net.add_layer(drop4, "dp4")
fc5 = FcLayer(num_hidden, num_output, params)
net.add_layer(fc5, "fc5")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1)
net.ShowLossHistory()
def net(init_method, activator):
max_epoch = 1
batch_size = 5
learning_rate = 0.02
eps = 0.01
params = HyperParameters41(learning_rate,
max_epoch,
batch_size,
eps,
net_type=NetType.Fitting,
init_method=init_method,
optimizer_name=OptimizerName.SGD)
loss_history = CLossHistory()
net = NeuralNet41(params)
num_hidden = [128, 128, 128, 128, 128, 128, 128]
count = len(num_hidden) - 1
layers = []
for i in range(count):
fc = FcLayer(num_hidden[i], num_hidden[i + 1], params)
net.add_layer(fc, "fc")
layers.append(fc)
ac = ActivatorLayer(activator)
net.add_layer(ac, "activator")
layers.append(ac)
# 从正态分布中取1000个样本，每个样本有num_hidden[0]个特征值
# 转置是为了可以和w1做矩阵乘法
x = np.random.randn(1000, num_hidden[0]).T
#x = np.random.normal(size=num_hidden[0]).T
# 激活函数输出值矩阵列表
a_value = []
a_value.append(x)
# 依次做所有层的前向计算
for i in range(len(layers)):
a = layers[i].forward(a_value[i])
a_value.append(a)
for i in range(count):
ax = plt.subplot(1, count, i + 1)
ax.set_title("layer" + str(i + 1))
plt.ylim(0, 10000)
if i > 0:
plt.yticks([])
ax.hist(a_value[i + 1].flatten(), bins=25, range=[0, 1])
#end for
# super title
plt.suptitle(init_method.name + " : " + activator.get_name())
plt.show()
def Net(subfolder,
dataReader,
num_input,
num_hidden,
num_output,
params,
show_history=True):
net = NeuralNet_4_2(params, subfolder)
fc1 = FcLayer_2_0(num_input, num_hidden, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Relu())
net.add_layer(relu1, "relu1")
fc2 = FcLayer_2_0(num_hidden, num_hidden, params)
net.add_layer(fc2, "fc2")
relu2 = ActivatorLayer(Relu())
net.add_layer(relu2, "relu2")
fc3 = FcLayer_2_0(num_hidden, num_hidden, params)
net.add_layer(fc3, "fc3")
relu3 = ActivatorLayer(Relu())
net.add_layer(relu3, "relu3")
fc4 = FcLayer_2_0(num_hidden, num_hidden, params)
net.add_layer(fc4, "fc4")
relu4 = ActivatorLayer(Relu())
net.add_layer(relu4, "relu4")
fc5 = FcLayer_2_0(num_hidden, num_output, params)
net.add_layer(fc5, "fc5")
softmax = ClassificationLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1, need_test=True)
if show_history:
net.ShowLossHistory(XCoordinate.Iteration)
return net
def Net(dataReader,
num_input,
num_hidden,
num_output,
params,
show_history=True):
net = NeuralNet41(params, "mnist_overfitting")
fc1 = FcLayer(num_input, num_hidden, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Relu())
net.add_layer(relu1, "relu1")
fc2 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc2, "fc2")
relu2 = ActivatorLayer(Relu())
net.add_layer(relu2, "relu2")
fc3 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc3, "fc3")
relu3 = ActivatorLayer(Relu())
net.add_layer(relu3, "relu3")
fc4 = FcLayer(num_hidden, num_hidden, params)
net.add_layer(fc4, "fc4")
relu4 = ActivatorLayer(Relu())
net.add_layer(relu4, "relu4")
fc5 = FcLayer(num_hidden, num_output, params)
net.add_layer(fc5, "fc5")
softmax = ClassificationLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1, need_test=True)
if show_history:
net.ShowLossHistory()
return net
def OverFitNet(num_input, num_hidden1, num_hidden2, num_output, params):
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid = ActivatorLayer(Relu())
net.add_layer(sigmoid, "sigmoid")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
tanh = ActivatorLayer(Relu())
net.add_layer(tanh, "tanh")
fc3 = FcLayer(num_hidden2, num_output, params)
net.add_layer(fc3, "fc3")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=10)
net.ShowLossHistory(0, None, 0, 1)
def model_sigmoid(num_input, num_hidden, num_output, hp):
net = NeuralNet(hp, "chinabank_sigmoid")
fc1 = FcLayer(num_input, num_hidden, hp)
net.add_layer(fc1, "fc1")
s1 = ActivatorLayer(Sigmoid())
net.add_layer(s1, "Sigmoid1")
fc2 = FcLayer(num_hidden, num_output, hp)
net.add_layer(fc2, "fc2")
softmax1 = ClassificationLayer(Softmax())
net.add_layer(softmax1, "softmax1")
net.train(dataReader, checkpoint=50, need_test=True)
net.ShowLossHistory("epoch")
ShowResult(net, hp.toString())
ShowData(dataReader)
num_input = dr.num_feature
num_hidden1 = 16
num_hidden2 = 4
num_output = 1
max_epoch = 1000
batch_size = 100
learning_rate = 0.1
eps = 0.001
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.MSE, InitialMethod.MSRA,
OptimizerName.Momentum)
net = NeuralNet(params, "PM25")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid1 = ActivatorLayer(Relu())
net.add_layer(sigmoid1, "sigmoid1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
sigmoid2 = ActivatorLayer(Relu())
net.add_layer(sigmoid2, "sigmoid2")
fc3 = FcLayer(num_hidden2, num_output, params)
net.add_layer(fc3, "fc3")
net.train(dr, checkpoint=10, need_test=True)
net.ShowLossHistory()
max_epoch = 50
batch_size = 32
learning_rate = 0.01
eps = 0.08
params = CParameters(
learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy3,
InitialMethod.MSRA,
OptimizerName.SGD)
net = NeuralNet(params, "Cifar10")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
r1 = ActivatorLayer(Relu())
net.add_layer(r1, "r1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
r2 = ActivatorLayer(Relu())
net.add_layer(r2, "r2")
fc3 = FcLayer(num_hidden2, num_hidden3, params)
net.add_layer(fc3, "fc3")
r3 = ActivatorLayer(Relu())
net.add_layer(r3, "r3")
fc4 = FcLayer(num_hidden3, num_hidden4, params)
net.add_layer(fc4, "fc4")
r4 = ActivatorLayer(Relu())
num_hidden = 32
num_output = 1
max_epoch = 1000
batch_size = 16
learning_rate = 0.1
eps = 0.001
params = CParameters(
learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy2,
InitialMethod.Xavier,
OptimizerName.SGD)
net = NeuralNet(params, "Income")
fc1 = FcLayer(num_input, num_hidden, params)
net.add_layer(fc1, "fc1")
sigmoid1 = ActivatorLayer(Sigmoid())
net.add_layer(sigmoid1, "sigmoid1")
fc2 = FcLayer(num_hidden, num_output, params)
net.add_layer(fc2, "fc2")
sigmoid1 = ActivatorLayer(Sigmoid())
net.add_layer(sigmoid1, "sigmoid1")
#net.load_parameters()
net.train(dr, checkpoint=10, need_test=True)
net.ShowLossHistory()
if __name__ == '__main__':
dataReader = LoadData()
num_input = dataReader.num_feature
num_hidden1 = 8
num_output = 3
max_epoch = 1000
batch_size = 10
learning_rate = 0.1
eps = 0.06
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy3, InitialMethod.Xavier,
OptimizerName.SGD)
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Relu())
net.add_layer(relu1, "relu1")
fc2 = FcLayer(num_hidden1, num_output, params)
net.add_layer(fc2, "fc2")
softmax1 = ClassificationLayer(Softmax())
net.add_layer(softmax1, "softmax1")
net.train(dataReader, checkpoint=10, need_test=False)
net.ShowLossHistory()
ShowResult(net, params.toString())
ShowData(dataReader)
def model():
dr = LoadData()
num_input = dr.num_feature
num_hidden1 = 32
num_hidden2 = 16
num_hidden3 = 8
num_hidden4 = 4
num_output = 1
max_epoch = 1000
batch_size = 16
learning_rate = 0.01
eps = 1e-6
params = HyperParameters(learning_rate,
max_epoch,
batch_size,
eps,
net_type=NetType.Fitting,
init_method=InitialMethod.Xavier)
net = NeuralNet(params, "HouseSingle")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
r1 = ActivatorLayer(Relu())
net.add_layer(r1, "r1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
r2 = ActivatorLayer(Relu())
net.add_layer(r2, "r2")
fc3 = FcLayer(num_hidden2, num_hidden3, params)
net.add_layer(fc3, "fc3")
r3 = ActivatorLayer(Relu())
net.add_layer(r3, "r3")
fc4 = FcLayer(num_hidden3, num_hidden4, params)
net.add_layer(fc4, "fc4")
r4 = ActivatorLayer(Relu())
net.add_layer(r4, "r4")
fc5 = FcLayer(num_hidden4, num_output, params)
net.add_layer(fc5, "fc5")
#ShowResult(net, dr)
net.load_parameters()
#Inference(net, dr)
#exit()
#ShowResult(net, dr)
net.train(dr, checkpoint=10, need_test=True)
output = net.inference(dr.XTest)
real_output = dr.DeNormalizeY(output)
mse = np.sum((dr.YTestRaw - real_output)**2) / dr.YTest.shape[0] / 10000
print("mse=", mse)
net.ShowLossHistory("epoch")
ShowResult(net, dr)
num_output = 10
max_epoch = 20
batch_size = 100
learning_rate = 0.2
eps = 0.08
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy3, InitialMethod.MSRA,
OptimizerName.SGD)
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid = ActivatorLayer(Relu())
net.add_layer(sigmoid, "sigmoid")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
tanh = ActivatorLayer(Relu())
net.add_layer(tanh, "tanh")
fc3 = FcLayer(num_hidden2, num_output, params)
net.add_layer(fc3, "fc3")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1, need_test=True)
def model_dropout():
dr = LoadData()
num_input = dr.num_feature
num_hidden1 = 64
num_hidden2 = 64
num_hidden3 = 64
num_hidden4 = 8
num_output = 1
max_epoch = 2000
batch_size = 16
learning_rate = 0.01
eps = 1e-6
params = CParameters(
learning_rate, max_epoch, batch_size, eps,
LossFunctionName.MSE,
InitialMethod.Xavier,
OptimizerName.Momentum,
RegularMethod.L1, 0.001)
net = NeuralNet(params, "HouseSingleDropout64")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
r1 = ActivatorLayer(Relu())
net.add_layer(r1, "r1")
#d1 = DropoutLayer(num_hidden1, 0.2)
#net.add_layer(d1, "d1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
r2 = ActivatorLayer(Relu())
net.add_layer(r2, "r2")
#d2 = DropoutLayer(num_hidden2, 0.3)
#net.add_layer(d2, "d2")
fc3 = FcLayer(num_hidden2, num_hidden3, params)
net.add_layer(fc3, "fc3")
r3 = ActivatorLayer(Relu())
net.add_layer(r3, "r3")
#d3 = DropoutLayer(num_hidden3, 0.2)
#net.add_layer(d3, "d3")
fc4 = FcLayer(num_hidden3, num_hidden4, params)
net.add_layer(fc4, "fc4")
r4 = ActivatorLayer(Relu())
net.add_layer(r4, "r4")
#d4 = DropoutLayer(num_hidden4, 0.1)
#net.add_layer(d4, "d4")
fc5 = FcLayer(num_hidden4, num_output, params)
net.add_layer(fc5, "fc5")
#ShowResult(net, dr)
net.load_parameters()
#Inference(net, dr)
#exit()
#ShowResult(net, dr)
net.train(dr, checkpoint=10, need_test=True)
output = net.inference(dr.XTest)
real_output = dr.DeNormalizeY(output)
mse = np.sum((dr.YTestRaw - real_output)**2)/dr.YTest.shape[0]/10000
print("mse=", mse)
net.ShowLossHistory()
ShowResult(net, dr)
#ShowData(dataReader)
num_input = dataReader.num_feature
num_hidden1 = 8
num_output = 3
max_epoch = 1000
batch_size = 10
learning_rate = 0.1
eps = 0.06
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy3,
InitialMethod.Xavier,
OptimizerName.SGD)
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
relu1 = ActivatorLayer(Sigmoid())
net.add_layer(relu1, "relu1")
fc2 = FcLayer(num_hidden1, num_output, params)
net.add_layer(fc2, "fc2")
softmax1 = ClassificationLayer(Softmax())
net.add_layer(softmax1, "softmax1")
net.train(dataReader, checkpoint=10, need_test=False)
net.ShowLossHistory()
ShowResult(net, params.toString())
ShowData(dataReader)
num_output = 10
max_epoch = 50
batch_size = 32
learning_rate = 0.01
eps = 1e-3
params = HyperParameters(
learning_rate, max_epoch, batch_size, eps,
net_type=NetType.MultipleClassifier,
init_method=InitialMethod.MSRA)
net = NeuralNet(params, "Cifar10")
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
r1 = ActivatorLayer(Relu())
net.add_layer(r1, "r1")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
r2 = ActivatorLayer(Relu())
net.add_layer(r2, "r2")
fc3 = FcLayer(num_hidden2, num_hidden3, params)
net.add_layer(fc3, "fc3")
r3 = ActivatorLayer(Relu())
net.add_layer(r3, "r3")
fc4 = FcLayer(num_hidden3, num_hidden4, params)
net.add_layer(fc4, "fc4")
r4 = ActivatorLayer(Relu())
num_output = 10
max_epoch = 20
batch_size = 100
learning_rate = 0.2
eps = 0.08
params = CParameters(learning_rate, max_epoch, batch_size, eps,
LossFunctionName.CrossEntropy3, InitialMethod.Xavier,
OptimizerName.SGD)
net = NeuralNet(params)
fc1 = FcLayer(num_input, num_hidden1, params)
net.add_layer(fc1, "fc1")
sigmoid = ActivatorLayer(Sigmoid())
net.add_layer(sigmoid, "sigmoid")
fc2 = FcLayer(num_hidden1, num_hidden2, params)
net.add_layer(fc2, "fc2")
tanh = ActivatorLayer(Tanh())
net.add_layer(tanh, "tanh")
fc3 = FcLayer(num_hidden2, num_output, params)
net.add_layer(fc3, "fc3")
softmax = ActivatorLayer(Softmax())
net.add_layer(softmax, "softmax")
net.train(dataReader, checkpoint=1, need_test=True)
